Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
MEKONG ECONOMIC RESEARH NETWORK
Final Draft
Determinants of Student Dropout in Cambodia’s Primary and Lower Secondary Schools: A Survey of Program Interventions
Ban Kosal Young Economist
Ministry of Economy and Finance (MEF)
Kim KinKesa Research Assistant
Centre for Policy Studies (CPS)
29 April 2015 Phnom Penh
1
Acknowledgments This work is carried out through a research grant and technical support from the Mekong Economic Research Network - a research initiative managed by the Centre for Analysis and Forecasting (CAF) of the Vietnam Academy of Social Sciences (VASS) with financial support from the International Development Research Centre (IDRC), Canada (project105220). The authors are grateful for helpful comments and suggestions by Mr. Chan Sophal, Cambodia National Coordinator of MERN.Views and errors in the paper are those of the authors only and do not necessarily represent those of any institution.
Abstract
This study finds out determinants of student dropout in both Cambodia’s primary and lower secondary schools. It also surveys program interventions used to mitigate dropout issue in other countries and Cambodia. Over ages students, working students and students from large family size are more likely to drop out of school. If student works for additional one hour from the average working hour, the probability of dropout increases by around 5% at primary level and 12% at lower secondary level. Working students tend to drop out of school when they graduate from primary school and are about to enter the lower secondary school. Share of household expenditure on education also plays important role to have an effect on student dropout. It is found that a higher share of household expenditure on education decreases the probability of school dropout. The policy implication from this finding is that if a scholarship is given directly to cover education expenditure, it will likely help reduce dropout headcount. Nevertheless, other interventions including school and teacher quality improvements should be also complementarily implemented to achieve quality education outcomes for students.
2
Determinants of Student Dropout in Cambodia’s Primary and Lower Secondary Schools: A Survey of Program
Interventions
1. Introduction
Cambodia considers its educational reform as one of the top priority policies. Since the first
election in 1993, gradual educational reforms have led to impressive enrolment and low
dropout rates for both genders at the primary level. These achievements are more
satisfactory than the target rates of the Education Strategic Plan (ESP) of the Ministry of
Education Youth and Sport (MoEYS). Nevertheless, dropout incidence at the lower
secondary level is still a significant concern for Cambodia’s education. In 2013, one in five
students dropped out of school after primary school completion, which is almost twice
higher than the target rate of 13% set by the ESP. Furthermore, this dropout rate was on an
upward trend from about 12% in 2003 to around 23% in 2011, with significantly higher
female dropouts causing another policy concern on gender disparity in education.
On the positive side, the last few years saw a slight decline of primary level dropout rate and
a steady lower secondary dropout rate. This favorable trend can be sustained with effective
policy interventions. The value of education can never be underestimated as Cambodia is
moving towards regional and global integration, particularly the 2015 ASEAN Economic
Community, when freer labor movements are expected to intensify. This study, therefore,
investigates the issues surrounding school dropout incidence at both primary and lower
secondary levels. It addresses several aspects of school dropout, using a mixed approach of
both qualitative and quantitative methodologies with available data from Cambodia Socio-
Economic Survey (CSES) and literature reviews on various types of program interventions.
Findings from this study could shed light on the significant policy options to tackle dropout
issue in Cambodia.
This paper is divided into 6 sections. The following section summarizes the literature review
while Section 3 gives an overview of scholarship support programs in Cambodia. Section 4
illustrates data and methodology. Section 5 covers the analysis and findings from CSES.
Finally, section 6 is the conclusion and policy recommendations.
3
2. Literature Review
This part summarizes and reviews factors contributing to dropout incidence and the
program interventions implemented by both government and development partners.
Particular focuses of this literature review are the compilation of outcomes and
effectiveness of each program intervention designed to retain students at school and
improve their study outcomes.
Factors Leading to Student Dropout
Education is an investment in human capital because of the expected benefits in return
(Mincer, 1958; Becker, 1964). People go to school because they believe in the dividend
return they will deserve for a better future life. Nonetheless, not all people successfully
complete their education. Some drop out from school for various reasons. Research studies
have found many factors affecting the school dropout. Those factors can be grouped as (1)
student factors, (2) family factors, (3) school factors, and (4) regional, community, country
factors. The student factors include the attribute factors such as gender, ethnic, racial, peer
context characteristics, student performance1, attitude, behaviours and background (Diyu,
2002; France, 2008; Heckman &LaFontaine, 2010; Rumberger, 2011; Jordan, Kostandini,
&Mykerezi, 2012). The family factors are involved with family size, poverty, parents’
education and so forth (Suliman& El-Kogali, 2002; Ana &Verner, 2006; Montmarquette,
Viennot-Briot, &Dagenais, 2007; Huisman& Smits, 2009). On the other hand, the school
factors could include school quality, curriculum, school regulation and teacher quality2
(McNeal, 1997; John, 1998; Huisman& Smits, 2009). The geographical factors include urban
and rural area, the distance from school to student’s house, local infrastructure and others
(Huisman& Smits, 2009; Jordan, Kostandini, &Mykerezi, 2012). Finally community, regional
and country level factors are related to the political stability, economics crisis, recession,
1Roderick (1993) and Rumberger (2001) show that students who experience academic difficulty had an elevated risk of dropping out. It is believed that dropping out from school can be explained as the culmination of a process of progressive disengagement with the academic and social dimensions. For some students the process starts much earlier event at elementary school (Rumberger 1987, 2000; Finn 1989; Newmann, Wehlage, and Lamborn 1992; Garnier, Stein, and Jacobs 1997). 2 In Cambodia, by using focus group discussion with parent, some studies found that because many parent think that teacher did not teach their children well. They think that it is a waste of time for their children because children are going to get nothing from the class. Therefore, parents decide to have their children dropped from school.
4
government supports and program on education, unemployment and others (Ravallion&
Quentin, 1999; Dre`ze&Kingdon, 2001; Huisman& Smits, 2009; Jordan, Kostandini,
&Mykerezi, 2012).
In Cambodia, there are some studies trying to empirically asses factors having impacts on
school dropout. Keng (2005) conducts field survey on household and children in two rural
villages of Pursat province in order to investigate the decision making on education when
the poor have to balance between future welfare and immediate needs. The results show
that in rural area, where parents are illiterate and children contribute to economic
production of the family, students are left to decide whether to contiue their education
after having enrolled in school for sometimes. Fata and Hirakawa (2012) also study factors
affecting school dropout in rural area. The survey was conducted in Kompong Cham
province through the stratified sampling on five primary schools and five lower secondary
schools based on dropout rate to examine the school factors. Totally 868 students from
first, fourth and seventh grade are included in the sample. The study shows that students
who are overage, have poor academic achievement, come from Champ ethnic and with
parents who have low aspiration on education, are likely to drop out. Bunchhay, Fata,
Sopha, and Hirakawa (2014), using similar method to Fata and Hirakawa (2012), they found
that teacher’s absence, mother’s education, repetition, parent living far away, and loss of
parent affect school dropout in rural primary school.
For this research study, differently it extends from previous researches on school dropout by
using the richer dataset of Cambodia Socio-Economic Survey (CSES) which covers 24
provinces across Cambodia. The immense size coverage of the CSES may give us better
generalized pictures i.e. the validity and the inference on characteristics of the school
dropout in Cambodia.
Program Interventions, Outcomes and Effectiveness
The literature reviews above reveal several factors leading to school dropout ranging from
student, to family, to community and to national level factors. For each factor, relevant
policy intervention could be carried out to address the school dropout accordingly. There is
a comprehensive study review on program intervention on school dropout through School
Dropout Prevention Pilot (SDPP) program, funded by USAID for four countries, namely
5
Cambodia, India, Tajikistan, and Timor Leste. Review findings show that in general the
program intervention to tackle school dropout can be categorized as follows (See Lorie,
Jennifer, &Rajani, 2011):
• Academic intervention: help student to achieve better school performance and
achievement. As indicated above, some of the dropout reasons could come from the
student performance and thus academic intervention is relevant to be done
• Financial intervention: help students and family to cope with school payment. As it is
widely believed, poor students are prone to dropout since they face not only direct
cost of education but also the opportunity cost (by working to earn income for the
family)
• Health intervention: target students whose poor health could lead to dropout
• Personal and Social Intervention: deal with attitude and behavior using approaches
such as group discussion, peer counseling and mentoring
• Structural intervention: target and connect community level, school level, provincial
level and national level together.
This report reveals that among these five interventions, financial intervention resulted in a
positive change in 12 of 13 cases (92%), followed by successful structural intervention with
three of four cases (75%). Academic intervention was successful in two of four cases (50%),
while health intervention had an impact on educational outcomes in two of five studies
(40%). Personal and social support were also successful, but it needs to be compelled with
other interventions at the same time. Besides this evaluation review, studies by (Ricardo,
Kwame, Jo, & Frances 2010) and (France 2008) also suggest similar categories of policy
intervention as those indicated by SDPP. Those interventions may be grouped as follows:
school related measures, financial and other measures (health intervention, community
involvement, adult education program, etc).
6
In addition, Shari et al (2013) also carry out the systematic review and impact evaluation
which analyses and synthesizes all relevant evidences about a specific intervention on
enrolment, dropout and education outcomes. They presented a new categorization of
supply and demand-side intervention and drew out lessons on each intervention type from
the 24 high-quality evaluation studies (see Figure 1).
Figure 1: Supply and Demand Side Approach to Education
Source: Shari et al (2013), page 04
Traditionally, education intervention has centered on supply-side of learning materials,
school building and teacher quality. More recently, increasing attention has also shifted to
the demand-side intervention. Demand-side intervention consists of three major categories,
namely (1) Reducing Cost (2) Providing Information and (3) Increasing Students’
Preparedness. There are different intervention types in the first and third categories.
Scholarship and conditional cash transfer support is grouped in the first category as it helps
ease education expenditure for students. The synthesis conclusions of the report can be
summarized as below (see Table 8in the Appendix on the overall pooled effect sizes by
outcome and intervention type):
7
• Scholarship/ conditional cash transfer could increase school enrolment and
attendance, but have no overall effect on students’ test scores. However, the
evidence base is not that broad for learning outcomes
• Enrolment and progress in school are improved by school fee subsidies, while merit-
based scholarships boost learning
• There is no effect on school attendance and language test scores of students in case
of distributing teaching and learning aids in school. Yet, it yields positive impacts on
mathematics test scores for computer-based learning and the regular school
curriculum interventions
• It looks promising in boosting schooling outcomes for interventions including teacher
investment, new schools, early childhood development program, community-based
school management and school-feeding programs.
3. Scholarship Support Programs in Cambodia
There are numerous scholarship supports in Cambodia, both in cash and in-kind payments.
The scholarships have ranged from government, development partners and individuals,
varying from a small number of students to the complete coverage. One major pioneering
project was funded by the Japan Fund for Poverty Reduction (JFPR) to provide scholarships
to lower secondary students. This scholarship program aimed at increasing student
enrollment and reducing the dropout numbers. The project was designed specifically for
female students from poor family and covered school year from 2003 to 2006.
The JFPR is a kind of Conditional Cash Transfer (CCT) program which has been widely
implemented and evaluated in many Latin American countries. In Cambodia, JFPR provided
scholarship of $45 each for female students at lower secondary schools. Female students
are automatically qualified for up to 3 years at their schools when they are awarded the
scholarships given the conditions that they maintain a passing grade are absent without
“good reason” less than 10 days in a year. However, there is no mechanism to ensure their
promising agreement is enforced. Scholarship recipients only agreed to use funds for their
education. The evaluation impacts of the JFPR show that it raises school attendance
8
approximately 30 percentage points higher than those without the project support. The
results also reveal the robust evidence of heterogeneous treatment effect in case of
Cambodia. Female students with low socio-economic status, low parental educational level
and living far away from schools see the largest positive effect on their enrollment and
attendance from this JFPR project (Deon Filmer & Norbert Schady, 2006).
Along with JFPR projects, there are also other scholarship support projects including the
Royal Government of Cambodia’s Priority Action Plan (PAP12); and the Basic Education and
Teaching Training (BETT) project of Belgian Technical Co-operation. The projects are very
similar in design and purposes. Drawing upon the lessons learnt of those projects, the
Cambodia Education Sector Support Project (CESSP) was supported and funded by the
World Bank. Among objectives to increase enrollment rate and reduce dropout rate of the
poor students at lower secondary schools, the project included the setting up of the
program called CESSP Scholarship Programme (CSP) with the initial budget of US$
5,840,256. The CSP covered five school years from September 2005 to July 2010 and
included two scholarship amounts i.e. $60 per year for the poorest recipients and $45 for
the others (Poisson, 2014). It is noteworthy that to ensure there is a single and consistent
MoEYS-run scholarship program, CSP took over the JFPR and BETT scholarship programs in
2007, when the World Bank was the only fund provider. However, the JFPR and BETT still
continued their funds to other education initiatives. To implement the CSP program, the
Local Management Committee (LMC) was established to select scholarship recipients and
administer the scholarship distribution (see Figure 2 below on CSP program administrative
structure).
Figure 2: CSP Program Administrative Structure
9
Source: (Poisson, 2014), page 107
The impact evaluation study of this CSP program by Marcus (2013) in the World Bank’s
policy note indicates that:
• The scholarships had a significant impact on student enrolment and attendance in
Grade seven and Grade eight. The impact was not different between girls and boys
• Providing students with the amount of $45 to stay in school proved as effective as
giving them the amount of $60
• The grants could be used on anything and how it was used was not tracked.
However, families that received funds spent more money on school related expenses
• Scholarship awardees were less likely to continue/leave for work while attending
school; and their siblings did not have to make up the difference in compensation for
families
• While funds increased enrolment among vulnerable students, it did not prove to
achieve better outcomes.
10
The final conclusion of the evaluation paper gives an additional suggestion that staying in
school is not always adequate to ensure a quality education. Besides the scholarship support
program of demand-side interventions, other interventions including school and teacher
quality improvements should be also complementarily implemented to achieve quality
education outcomes for students.
There is also another scholarship program supported by GPE (Global Partnership for
Education). It provided scholarship with the amounts of $2,100,000 per fiscal year 2014-
2015. About 70,000 poor and good performing students received US$ 30 per year. Most of
them are selected based on the Ministry of Planning’s ID Poor 1 and ID Poor 2. The coverage
is 24 provinces, particularly those who are not covered by other scholarships. However,
from school year 2015-2016, the Cambodian government will cover the scholarships with its
own budget. A number of draft documents3 suggest that to upcoming government
scholarship program to be finalized by 2015 will extend the scholarship from Grade four in
the primary school to Grade nine in the upper secondary school so as to reduce the high
dropout incidence during the student transition from the primary to lower secondary
school.
In case of the primary school, on the other hand, two cash and in-kind programs are
supported by the World Food Programme (WFP) and government:
• School feeding program: breakfast for all students at selected primary schools from
Grade one to Grade six including kindergarten located in those primary schools. The
program runs for 10 months/year and covers 385,000 students in 12 provinces.
• Scholarship (10kg of rice/months OR 20,000 Riels/month) for 10 months/year in 15
provinces. This is done only for ID Poor 1 and those qualified similar to ID Poor 1
based on identification tools adapted from Ministry of Planning and following
confirmation with community and schools. It is being delivered to 97,000 students in
Grade four to Grade six.
3This scholarship information was drawn upon the informal consultation with MoEYS officials in charge of scholarship support program. However, the finalized scholarship scheme and its coverage might be subject to change.
11
It is noted that the Cambodian government contributes around 2,000 tonnes of rice to WFP
for these programs annually.
4. Data Sources and Methodology
This study mainly uses the Cambodia Socio-Economy Survey (CSES), which is conducted by
the national institute of statistics (NIS) of the Ministry of Planning. In this study, CSES 2003,
CSES 2007, CSES 2009, CSES 2011, and CSES 2012 are used to analyze the dropout in
Cambodia since they have tractable link and consistent questionnaires, allowing for the
comparison of school dropout over the years. CSES before 2003 are excluded from the
analysis because they do not have consistent questionnaires with CSES after 2003.In
addition, other sources are also used to retrieve relevant information for the analysis such
as Education Management Information System (EMIS) data, data from international
institutions such as UNICEF, UNESCO, World Bank and others.
For the methodology, Probit and Logitregression methods are used to investigate the
characteristics of dropout and its determinants. Three set of variables are controlled in the
regressions: (1) individual characteristics, (2) family characteristics and (3) school and
community contextual factors. The regressions are applied separately to primary school’s
dropout group and lower secondary school’s dropout group, and further classified into
Cambodia, Urban, and Rural Areas.
Following Ana and Verner (2006), the model is used to estimate the effect of each
determinant on the probability of dropout with the following specification.
𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ = 𝑿𝑿𝒊𝒊𝜷𝜷 + 𝑜𝑜𝑖𝑖 (1)
Where 𝑿𝑿𝒊𝒊 is a vector of explanatory variables including individual characteristic, family and
community contextual factors and 𝑜𝑜𝑖𝑖 is the error term (uncontrolled factor that can
affect (𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗). We do not observe 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ but we observe the child actually
drops out if 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ ≥ 0 and a child is still at school if 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ < 0, we can
write:
𝑑𝑑𝑑𝑑𝑜𝑜𝑑𝑑𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖 = �1, 𝑖𝑖𝑜𝑜 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ ≥ 00, 𝑖𝑖𝑜𝑜 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ < 0 (2)
12
It is possible to regard 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ as the net benefits or net utility of dropping out of
school (benefits after offsetting the cost of dropout). We do not know exactly the
magnitude of the net benefit, but we know that if the net benefit of dropout from school is
positive (𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ ≥ 0), the child decides to drop out, and we observe that the child is
currently a dropout. However, if the net benefit of dropout from school is
negative (𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ < 0), the child decides to stay in school, and we observe that the
child is currently in school (despite the child is a part-time worker or having regular
absence). Because 𝑑𝑑𝑑𝑑𝑜𝑜𝑑𝑑𝑜𝑜𝑜𝑜𝑜𝑜 takes value 1 (dropout) and 0 (non-dropout), the probability
model fits well for the estimation. The probability of student dropout can be written as
follows:
𝑃𝑃(𝑑𝑑𝑑𝑑𝑜𝑜𝑑𝑑𝑜𝑜𝑜𝑜𝑜𝑜 = 1) = 𝑃𝑃(𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑖𝑖∗ ≥ 0)
= 𝑃𝑃(𝑿𝑿𝒊𝒊𝜷𝜷 + 𝑜𝑜𝑖𝑖 ≥ 0)
= 𝑃𝑃(−𝑜𝑜𝑖𝑖 < 𝑿𝑿𝒊𝒊𝜷𝜷)
𝑷𝑷(𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅 = 𝟏𝟏) = 𝑭𝑭(𝑿𝑿𝒊𝒊𝜷𝜷)(3)
Where 𝑭𝑭( . )is a cumulative density function of 𝑜𝑜𝑖𝑖, the Probit model is used if the error term
𝑜𝑜𝑖𝑖follows a cumulative standard normal density function, and the Logit model is used if the
error term 𝑜𝑜𝑖𝑖 follows a logistic distribution function. We estimate equation (3) with both
Probit and Logit models.
5. Analysis and Findings
Descriptive Statistics on Dropout
Using the information from CSES, the dropout is defined as students who used to be in
school, that is ever attended school and who is not currently in school system. This
13
definition distinguishes dropouts from children who never attend school. This definition is
consistent with UNESCO’s and MoEYS’s definitions. Only individuals whose ages are equal to
or less than 16 at the time of interview are counted as dropouts since we focus on dropout
incidence at primary and lower secondary school. In Cambodia, students who go to school
at the age of 6 are expected to finish primary school at age 12 and finish lower secondary
school at age 15 or 16. Moreover, if a child is on holiday, he or she is considered as in the
school system.
Figure 3: Dropout Rate at Primary School from 2003-2012
Figure 4: Dropout Rate at Lower Secondary School from 2003-2012
Source: Author’s calculation from CSES 2003-2012
Figure 3 and Figure 4 display the estimated percentage of dropouts from 2003 to 2012 at
primary and lower secondary school. The dropout rates for female students are higher in
6.0 6.87.9 8.0 7.9
5.6 5.7
0%
2%
4%
6%
8%
10%
2003 2007 2008 2009 2010 2011 2012
total male female
12% 11%15%
18% 18%23% 22%
0%
5%
10%
15%
20%
25%
30%
2003 2007 2008 2009 2010 2011 2012
14
both levels. The trend of dropout rates in the last several years reveals a slight decrease in
overall dropout rate at the primary level, but not at the lower secondary level, which stays
at 22% in 2012 compared to 5.7% at the primary level. This favorable trend may continue
and spill over the lower secondary level if effective policy interventions are introduced and
carried out.
The dropout rate by class in academic year 2012-2013 is displayed in Figure 5. The graph
illustrates that the dropout rate reaches the peak at the grade 5 of primary level and is
around 20% for the lower secondary grades. The high rates at the lower secondary level
could be explained by two main reasons. Firstly, after the completion of primary school, on
average students reach the age of 12, which for poor families, the students are able to work
and support their families. This is especially true for overaged children. Secondly, because of
a lack of lower secondary schools and a long distance from their villages, many students
decide to give up on their lower secondary education before or after the completion of
primary school. These cases tend to happen frequently in many rural areas of Cambodia
(Rajani, Jennifer, & Karen, 2011). Similarly, the dropout rate at grade 9, which is a transition
grade to upper secondary level, is also higher than that of grade 8.
High dropout rates at the end of primary and lower secondary levels are common in other
poor countries. Lewin (2007) and Ricardo, Kwame, Jo, & Frances (2010) found that countries
in Sub-Saharan Africa such as Uganda, Rwanda, Cameroon and Kenya have a high dropout
rate at the end of primary and lower secondary school cycles.
Figure 5: Dropout Rate by Class 1 to 9
Source: Author’s calculation from EMIS, academic year 2012-2013
7.8%5.3%
7.3%
12.1%
19.5%
11.7%
20.9% 19.7%
23.4%
0%
5%
10%
15%
20%
25%
1 2 3 4 5 6 7 8 9Grade
15
By province, the statistics from EMIS shows that dropout rates in Cambodia are relatively
high in coastal and plateau regions (see Figure 6 and Figure 7). At the primal level, in
academic year 2012-2013, eight provinces had two digits dropout rates including Koh Kong,
Stung Treng, and RatanakKiri, which were top three with more than 15% dropout rates. In
general, female dropout rates were higher in most provinces, regardless of their
geographical locations, except in Koh Kong, Stung Treng, OtdarMeanchey, and Preah
Sihanouk, where male dropout rates were much higher than female counterparts. Koh Kong
province surprisingly had the lowest dropout rate at the lower secondary level. The top
three provinces that had the highest dropout rates (more than 25%) were OtdarMeanchey,
BanteayMeachey, and Kampong Speu.
Figure 6: Dropout Rate at Primary Level by Province (%)
Source: Author’s calculation from EMIS, academic year 2012-2013
-
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
Total Female Male
16
Figure 7: Dropout Rate at Lower Secondary Level by Province (%)
Source: Author’s calculation from EMIS, academic year 2012-2013
In general, the dropout rates in urban areas are lower than that in rural areas. As shown in
Figure 8, the dropout rate at primary level in urban area is only 7.8% compared to 10.9% in
rural area. Similarly, the dropout rate at lower secondary school in urban area is only 14.3%
compared to 23.2% in rural area. Such geographical difference may be explained by
difference in number of the poor where a large number of them tend to live in rural area. It
also reflects the difference in supply constraint because available schools in rural area tend
to locate farer from home than school in urban area (see Table 7for the distance
comparison in the Appendix).
10.0
15.0
20.0
25.0
30.0
35.0
Total Female Male
17
Figure 8: Dropout Rate in Urban and Rural Areas
Source: Author’s compilation from EMIS
A list of factors given by respondents as the reasons for dropout is shown in Figure 9. The
most contributed factors leading to school dropout based on the answers of respondents at
both primary and lower secondary levels are “must contribute to household’s income”,
“must help with household chores”, “don’t want to study” and “did not do well in school”.
All these factors account for about 80% of the reasons, which shows that household and
individual factors are main determinants of school dropout in Cambodia.
Figure 9: Reasons of School Dropout
7.8%
10.9% 10.5%
14.3%
23.2%21.2%
0%
5%
10%
15%
20%
25%
Urban Rural Total Urban Rural Total
Primary Lower Secondary
18
HH Income
23%
HH Chores21%No
Interests18%
Low Perferma
nce15%
Too Poor10%
Other13%
Primary Level
HH Income
34%
HH Chores21%
No Interests
13%Low
Perfermance14%
Too Poor9%
Other9%
Lower Secondary Level
Source: Author’s calculation from CSES 2003-2012
Even though the statistics from the survey shows some indicators affecting school dropout,
it is not enough for policy implications because there are two limitations. Firstly, it is hard to
draw the cause and effect conclusion because these factors could jointly affect or interact
on dropout’s decision. In this paper, the probabilistic regression method will be used to
address the causes of school dropout at both primary and lower secondary levels. Secondly,
the relative effect of each factor is not clear. For instance, it is significant to investigate if an
income increase of 1% can reduce the probability of dropout. This question can be
addressed with the regression method.
Table 1 shows the description and definitions of the independent variables to be used in this
analysis. The variable “sex” is defined by female students equal to “0” and male equal to
“1”. The variable “age” is the age of students; “famsize” is the number of members in a
household; “headedu” is the education level of household head. The variable “lnincome” is
natural logarithm of monthly household income in Riels; “eduratio” is the share of
education expenditure in annual household total expenditure; and finally “lnworkhrs” is the
natural logarithm of working hours in the past 7 days.
Table 1: Variable Definitions
Variable Description Note
sex Sex 0 = Female, 1 = Male
age Age
famsize Family size Number of Family Members
headedu Household head’s education
lnincome Monthly household income (Riels) Natural Logarithm
eduratio Share of education in annual household expenditure Percentage Point
lnworkhrs Working hours in the past week Natural Logarithm
A comparison of statistics between dropout and non-dropout is shown in Table 2. The table
reveals that dropouts at both primary and lower secondary schools tend to have older ages,
19
a larger family, and a household head with a lower education. Moreover, they tend to have
a lower income and own less valuable assets, particularly land. The dropouts live in a family
who on average owns about 1.5 ha, which is smaller compared to the land owned by an
average family of non-dropouts, which is 2.7 ha at primary level and 1.9 ha at lower
secondary level.
Table 2: Differences between Dropout and Non-dropout
age famsize fathedu mothedu headedu totincome land size
Primary
Non-dropout 11.0 6.1 7.0 5.8 6.9 1,810,000 27,976
Dropout 13.6 6.3 5.4 4.6 5.3 1,350,000 15,168
Total 11.1 6.1 6.9 5.7 6.8 1,780,000 27,129
Urban
Non-dropout 10.7 6.0 8.7 7.0 8.2 2,660,000 16,902
Dropout 13.4 6.3 5.6 3.9 5.7 1,660,000 16,323
Total 10.8 6.0 8.6 6.9 8.1 2,590,000 16,865
Rural
Non-dropout 11.1 6.2 6.5 5.3 6.4 1,450,000 29,810
Dropout 13.7 6.3 5.3 4.8 5.3 1,240,000 14,988
Total 11.2 6.2 6.4 5.3 6.4 1,430,000 28,822
Lower Secondary
Non-dropout 14.6 6.0 7.9 6.5 7.8 2,460,000 19,503
Dropout 15.2 6.2 5.9 4.9 5.9 1,390,000 15,169
Total 14.7 6.0 7.6 6.2 7.5 2,290,000 18,754
Urban
Non-dropout 14.4 5.9 9.1 7.2 8.8 2,590,000 22,987
Dropout 15.3 5.9 6.3 5.1 6.2 1,310,000 18,287
Total 14.5 5.9 8.9 7.0 8.5 2,440,000 22,386
Rural
Non-dropout 14.7 6.1 7.0 5.9 7.0 2,350,000 18,277
Dropout 15.2 6.3 5.8 4.8 5.8 1,430,000 14,474
Total 14.8 6.1 6.8 5.7 6.8 2,170,000 17,563
Source: Author’s calculation from CSES 2003-2012
Regression Results and Discussions
20
To estimate the determinants of school dropout, this study applies Probit model regressions
by classifying students into urban, rural and all areas. There are several grounds to support
his classification. Firstly, students who live in urban areas statistically have a lower
probability of dropout at both primary and lower secondary levels, as shown in the
descriptive statistics section. In rural areas, especially in remote area where poverty trap is
common, the need for daily survival reduces incentives of many children to pursue their
study to a higher level. In addition, the number of schools, especially at lower secondary
level, is substantially lower in rural areas. Secondary schools are usually located in the
provincial or district towns far away from villages. As a result, many children drop out of
school after finishing primary education in their village schools.
It should be noted that the sample size for the urban areas are substantially smaller than
other areas after the classification. The small sample size slightly affects the significant levels
and marginal effects of each coefficient in the primary level model, but significantly biases
(underestimates the coefficient and overestimates the standard errors of) those statistics in
the urban area model of the lower secondary level regressions as shown below.
Primary Level
At primary level, Table 3 shows the results of Probit model, and Table 4 shows the marginal
effects of each coefficient. Probit and Logit models provide similar results4. The regression
results show that male students are less likely to drop out of school than female students.
As shown in Table 4, male students in urban areas have a dropout probability of 18.6%
lower than female students although the coefficients are not statistically significant in
general and in the rural areas.
Obviously, the potential of over age can increase the probability of dropout at primary level.
The regression coefficient is positive and significant, showing a positive relationship
between age and dropout probability in all three models. If the age of students increases by
one year, from the average age, the probability of dropout increases by 5.2%, 7.7% and
4 The results of Logit model are not shown in this study, but available upon request. Particularly in Column 1 of Table 7, the pseudo R2 is the same with that of the Logit model (pseudo R2 =0.30). Also, both Probit and Logit models’ goodness-of-fits indicate that both models can correctly predict 81% of the observations However, the log likelihood of Probit model is slightly lower than that of Logit model (-158.50 versus -157.09). Therefore, we interpret the result using Probit model. However, the interpretation of Logit model is the same in term of the signs of coefficients, despite slightly different magnitudes of coefficients.
21
4.6% in All, Urban, and Rural areas respectively. Theoretically, older students have more
potential to work than their younger counterparts, and, thus, are more likely to drop out.
This phenomenon is especially more severe in the urban area, where job opportunities are
more available than in the rural areas.
Family size is also an important determinant of school dropout. From the table, the
coefficients in the regressions are positive and significant for both rural and all areas, which
suggests that a larger family can raise the dropout probability of children, especially older
children. Usually, for a poor family with a large number of children, the elder are prone to
earn income to support family and provide for the younger to continue education.
Moreover, a family with a large number of children also incurs a large cost of education if all
children are allowed to study. According to the marginal effect in Table 4, one additional
member increases the probability of dropout by around 1% for all, urban and rural areas.
Education of the household head has been found to play an important role on the education
of young members in the family, as discussed in the literature. Generally, many studies have
argued that a parent’s education strongly affects children’s education. It is believed that if
parents are highly educated, the children are likely encouraged to continue their study to a
higher education. One explanation is that parents with a high education tend to appreciate
education and, hence, are willing to encourage their children to attain more education. In
Cambodia, this is true not only for children that are currently living with both parents, but
also with a single household head (with only father or mother) and with a household head
as a grandparent, an uncle, an aunt, a brother, a sister and so on. This determinant is
particularly important at the lower secondary level. Empirically, the coefficient of household
head’s education is statistically significant at lower secondary school, but not at the primary
level. This reflects the fact that household head’s education matters most during the
decision to pursue a higher education, particularly from primary to lower secondary
education.
Surprisingly, the regression results for both the primary and lower secondary levels show
that although the level of household incomes is inversely related to the probability of
dropout, this study fails to find any significant effects of this variable. One of the reasons for
this insignificant effect is due to the deficiency of income (or expenditure) data reported by
the respondents in CSES survey. It is quite common that households underestimate their
22
income and overestimate their expenditure during the survey interviews, apparently to
withhold their economic status for fear of crimes (in relatively rich households) and
expectation of financial supports (in relatively poor households).
To capture the impact of household’s education expenditure share on school dropout
decision, this study includes the education share of total household expenditure as one
determinant. The result shows that if a household spends more on education, the
probability of dropout is lower. One percent increase in education share of the household
reduces the probability of dropout by around 15% in all, urban, and rural areas, which is a
tremendous effect. This is also another explanation for the minimal effect of income level
on dropout probability. Only the increase in education expenditure share can effectively
reduce dropout probability. This is important for policy makers in terms of financial
interventions such as a scholarship that directly supports expenditure on education, rather
than goes directly into the total household income. The latter channel may have a small
impact on the probability of dropout because a household is more likely to use that support
for other purposes rather than for education. However, a scholarship that goes directly into
expenditure share for education is more effective in reducing dropout.
Finally, to account for the opportunity cost of education, the model controls for the working
hours of children. As suggested by many previous studies, the working hours will have a
positive effect on school attendance and dropout (UNICEF, ILO, & WB, 2006; Guarcello,
Lyon, &Rosati, 2004). However, there is a debate on the threshold level of working hours,
beyond which students will prefer to work rather than to attend school. In Cambodia some
research has found that there is such a threshold level of working hours that affects school
dropout (Phoumin& Fukui, 2006; Peter, 2008). Although our result cannot identify the
threshold effect, it is shown that the coefficient of working hours is positive and very
significant, which shows the positive relationship between working hours and dropout
probability. The marginal effect in Table 4 reveals that an hour increase in working hours
from its mean value increases the probability of dropout by 4.7%, 10.8%, and 4.0% for all,
urban, and rural areas respectively. This result reveal a significant finding on the role of
opportunity cost for primary education in Cambodia, especially in the urban areas where job
opportunities are more available.
23
Table 3: Determinants of Dropout Probability at Primary Level
(1) (2) (3)
All Urban Rural
sex -0.106 -0.792*** -0.000
(-0.996) (-2.776) (-0.003)
age 0.338*** 0.334*** 0.337***
(9.645) (3.685) (8.683)
famsize 0.069*** 0.046 0.058**
(2.637) (0.731) (1.980)
headedu -0.016 -0.023 -0.016
(-0.890) (-0.564) (-0.778)
lnincome -0.008 -0.092 0.003
(-0.184) (-0.755) (0.051)
eduratio -1.017* -0.562 -1.275*
(-1.838) (-0.505) (-1.924)
lnworkhrs 0.843*** 1.282*** 0.795***
(8.875) (4.861) (7.647)
_cons -8.213*** -7.324*** -8.274***
(-8.669) (-2.860) (-7.898)
N 1101 162 939
pseudo R2 0.300 0.419 0.288
AIC 733.451 127.287 600.953
ll -358.725 -55.644 -292.477
chi2 307.773 80.146 236.622
p 0.000 0.000 0.000
Note: t statistics in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01
24
Table 4: Marginal Effects of the Explanatory Variables at Their Mean Values
All Urban Rural
Variable dy/dx dy/dx dy/dx
sex -1.6% -18.6% 0.0%
age 5.2% 7.7% 4.6%
famsize 1.0% 1.1% 0.8%
headedu -0.3% -0.5% -0.2%
lnincome -0.1% -2.1% 0.0%
eduratio -15.5% -12.9% -17.6%
workhrs 4.7% 10.8% 4.0%
Lower Secondary Level
Table 5 shows the results of Probit model, and Table 6 shows the marginal effects of each
coefficient for the lower secondary level. As discussed above, we take precautions in
interpreting the results in the urban area because the sample size is a bit small, only 88
observations.
Unlike the primary level, the regression results show that sex is not a significant
determinant of student dropout at the lower secondary level. However, Table 5shows an
interesting contrast between roles of gender in the urban and rural areas. In general and in
the rural areas, male students have a lower dropout probability than female students do
although the coefficients are not statistically significant, yet the former seem more likely to
drop out than the latter in the urban area. Over age can also increase the probability of
dropout. Similar to the primary level, if the age of students increases by one year, from the
average age, the probability of dropout increases by 5.3% and 5.0% in all and rural areas,
respectively, although the marginal effect in the urban area is minimal probably due to the
25
sample issue. Older students, at the lower secondary level, have much more potential to
work than their younger counterparts, and, thus, are more likely to drop out.
Family size is again an important determinant of school dropout at the lower secondary
level. From the table, the coefficients in the regressions are positive and significant for both
rural and all areas, which suggests that a larger family can raise the dropout probability of
children. According to the marginal effect in Table 6, this variable has a stronger impact on
dropout probability than at the primary level. As discussed above, the burdens faced by
households at the primary level are even heavier for households at the lower secondary
level. One additional member has been found to increase the probability of dropout by
around 2.5% and 3.7% for students in all and rural areas, which comparatively much higher
in the case of primary level.
Education of the household head has been found to play an important role on the education
of young members in the family, as discussed. This determinant is more pronounced at the
lower secondary level. The coefficient of household head’s education is statistically
significant at lower secondary school in all, urban and rural areas, but not significant at the
primary level. This reflects the fact that household head’s education matters most during
the decision to pursue a higher education, particularly from primary to lower secondary
education.
Reinforcing the results at the primary level, the result shows one percent increase in
education share substantially reduces dropout probability by around 40% in all and rural
areas at the lower secondary level. Finally, it is shown that the coefficient of working hours
is positive and very significant. The marginal effect reveals that an hour increase in working
hours at the lower secondary level increases dropout probability of by 12.2%, 0.8%, and
11.6% for all, urban, and rural areas respectively. This result is even more alarming than at
the primary level and shows that the role of opportunity cost for lower secondary education
in Cambodia is non-negligible.
26
Table 5: Determinants of Dropout Probability at Lower Secondary Level
(1) (2) (3)
All Urban Rural
sex -0.020 0.410 -0.102
(-0.123) (0.692) (-0.573)
age 0.198** 0.424 0.164**
(2.569) (1.292) (2.004)
famsize 0.094** -0.151 0.121***
(2.268) (-0.793) (2.690)
headedu -0.101*** -0.238* -0.093***
(-3.477) (-1.905) (-2.860)
lnincome -0.002 -0.272 0.022
(-0.023) (-1.059) (0.306)
eduratio -1.492** -1.364 -1.362**
(-2.342) (-0.583) (-1.978)
lnworkhrs 1.249*** 3.675*** 1.040***
(8.308) (3.686) (6.591)
_cons -7.502*** -13.029* -6.856***
(-4.311) (-1.744) (-3.683)
N 377 88 289
pseudo R2 0.301 0.708 0.242
27
AIC 333.015 44.285 285.276
ll -158.508 -14.142 -134.638
chi2 136.415 68.428 85.979
p 0.000 0.000 0.000
Note: t statistics in parentheses, * p< 0.10, ** p< 0.05, *** p< 0.01
Table 6: Marginal Effects of the Explanatory Variables at Their Mean Values
All Urban Rural
Variable dy/dx dy/dx dy/dx
sex -0.5% 0.3% -3.1%
age 5.3% 0.2% 5.0%
famsize 2.5% -0.1% 3.7%
headedu -2.7% -0.1% -2.8%
lnincome 0.0% -0.2% 0.7%
eduratio -39.6% -0.8% -41.1%
workhrs 12.2% 0.8% 11.6%
6. Conclusion and Policy Recommendation
The study empirically found that main factors of school dropouts in Cambodia are closely
related to student and, more importantly, family characteristics. For student factors, over
aged children are more likely to drop out of school. From the analysis, if age increases by
one year from the average age, the probability of dropout increases by around 5%. Older
28
students tend to have a higher opportunity cost than their younger counterparts because
their labor potential is an important pulling factor, especially for poor families. From our
empirical analysis, opportunity cost significantly affects school dropout probability because
working children are more likely to leave school than non-working children. Working
children tend to drop out of school when they graduate from primary school and are about
to enter a lower secondary school. Gender disparity is also a concern based on our empirical
result. Although it is not significant, female students have a slightly higher probability of
school dropout than male students.
Family characteristics play an important role in students’ decision to drop out. Firstly,
students with a larger family size are more likely to drop out than those with a smaller
family size. In a poor family with many members, the elder children are usually expected to
earn income to support their family and younger ones to continue their education.
Moreover, families with a large number of children also incur a large cost of education if all
children were allowed to study. Secondly, students whose parents or household heads have
a low education are more probable to leave school. One explanation is that parents with a
high education tend to value education and, hence, encourage their children to attain a
higher education, and vice versa.
Last family factor that affects students’ decision is the share of household expenditure on
education. A higher share of household’s expenditure on education reduces the probability
of school dropout. Households usually spend more on other purposes rather than education
when their income increases. Interestingly, the level of household income does not have a
significant impact on students’ decision. This suggests that financial intervention such as a
scholarship that directly goes into overall income may has little impact on the probability of
dropout because of the fungibility issue; households are more likely to use the new income
source for other purposes rather than for education. The policy implication from this finding
is that if a scholarship is given directly to cover education expenditure, it will likely help in
reducing dropout.
Stemming from the above findings, the following policy recommendations are suggested to
tackle the dropout issue in Cambodia:
29
1. Policies with financial, mental, consulting supports for student groups with high
probability of dropout and their families are needed. These groups include:
Students live in a poor family Working students, especially students whose family is poor and family
size is large Students live in rural areas, particularly female students Students whose parents do not well understand the importance of
education. • Public awareness on benefit of education for uneducated and less
educated parents is needed.
2. More incentives and financial supports to teachers, students, and school staffs
for a visit to convince families of the dropout back to school will help reduce
dropout rate. This is consistent with recommendationsby ADB and MoEYS on
School Based Enrichment Program (MoEYS and ADB, 2012).
3. Education expenditure still has a strong effect on family decision relating to
children education, mainly poor families. Financial supports such as an increase
of scholarship allowances to students with a high risk of dropout will be effective
although the efficient amounts should vary between primary and lower
secondary levels. However, it is important to ensure that the purpose of
scholarship is toward education.
4. Encourage the integration of dropout issues into local authority development
plan with clear targeted goals such as development and budget plans of the
Sangkat or Commune.
5. Improve coordination between local authorities, schools, and families to jointly
solve dropout issues; and public awareness on dropout issue should target areas
with high rates of lower secondary school dropout.
6. Improving school environment, curriculum quality, and teacher quality are
always needed and are a continuing process to encourage students staying at
30
school. Curriculum should be updated in response to market and social demand
so that parents will value the future benefits of education.
Reference
Ana, R. C., & Verner, D. (2006). School Drop-out and Push-out Factors in Brazil: The Role of Early Parenthood, Child Labor, and Poverty. IZA Discussion Paper, 2515.
Becker, G. S. (1964). Human Capital. New York: Columbia University Press.
Bunchhay, A., Fata, N., Sopha, S., & Hirakawa, Y. (2014). Dropout In Rural Primary School In Cambodia: A Multilevel Analysis. Comparative and International Education Society Annual Conference 2014. Toronto.
Cameron, A. C., & Trivedi, K. P. (2009). Microeconometrics Using Stata. Texas, United State of America: Stata Press.
Christine, A. K., Micheal, & Nelson. (2007). School characterisics related to high school dropout rates. Remedial and Special Education, 28 (6), 325-339.
Colclough, C., & Lewin, K. (1993). Educating All the Children: Strategies for Primary Education in Developing Countries. Oxford: Oxford University Press.
31
Debbie, B., & Jennifer, E. C. (2004). High School Dropouts:Can We Reverse the Stagnation in School Graduation? Issue Brief, 1 (2), 1-11.
Department of Public Instruction: Safe and Healthy School Support Division. (2014). Dropout Data Collecting and Reporting Procedures Manual March 2014. North Carolina.
Diyu, X. (2002). Investigation and discussion on the problem of primary and secondary school dropout in poor areas. Chinese Education and Society, 33 (5), 49-58.
Dre`ze, J., & Kingdon, G. (2001). School Participation in Rural India. Review of Development Economics, 5 (1), 1-24.
Fata, N., & Hirakawa, Y. (2012). Identifying causes of drop-out through logitudinal quantitative analysis in rural Cambodia basic schools. Journal of International Development and Cooperation, 19 (1), 25-39.
France, H. (2008). Dropping Out from School: A Cross-Country Review of Literature. Research Monograph, No 16, pp. 1-67.
Guarcello, L., Lyon, S., & Rosati, F. (2004). Impact of Working Time on Children’s Health. Understanding Children's Work: An Inte-Agency Research Cooperation Project .
Hammack, F. M. (1986). Large School System's Dropout Reports: An Analysis of Definitions, Procedures, and Findings. Teachers College Record, 87 (3), 324-341.
Hayes, R., Nelson, J., Tabin, M., Pearson, G., & Worthy, C. (2002). Using school-wide data to advocate for student success. Professional School Counseling, 6 (2), 86-95.
Heckman, J. J., & LaFontaine, P. A. (2010). The American high school graduation rate: Trends and levels. The Review of Economics and Statistics, 93 (2), 244-262.
Holmes, J. (2003). Measuring the determinants of school completion in Pakistan: Analysis of censoring and selection bias. Economics of Education Review, 22 (3), 249-261.
Huisman, J., & Smits, J. (2009). Keeping Children in School: Household and District-level Determinants of School Dropout in 363 Districts of 30 Developing Countries. NiCE Working Paper 09-105 , pp. 1-38.
John, W. (1998). The Relationship of School and Cumminity Characteristics to High School Drop-Out Rate. The Clearing Hourse, 71 (3), 184-188.
32
Jordan, J. L., Kostandini, G., & Mykerezi, E. (2012). Rural and urban high school dropout rates: Are they different? Journal of Research in Rural Education, 27 (2), 1-21.
Kampuchean Action for Primary Education (KAPE). (2013). Enrollement Trend in Phnom Penh: A Need Assesment. Phnom Penh: Save the Children International.
Keng, C. (2005). Decision-making on Education: The role of children. Evidence from Rural Cambodia. Journal of Southeast Asian Education, 5 (1&2), 20-32.
Kitamura, Y. (2008). Basic Education Reforms in Cambodia and Bangladesh:Expansion of Acess and Improvement of Quality. Forum of International Development Studies , 85-102.
Lewin, K. (2007). Improving Access, Equity and Transitions in Education: Creating a Research Agenda. Brighton: University of Sussex.
Lorie, B., Jennifer, S., & Rajani, S. (2011). Inventory of Policies and Programs Related to Dropouts in Cambodia, India, Tajikistan, and Timor Leste. USAID.
Lyche, C. (2010). Taking on the Completion Challenge: A Literature Review on Policies to Prevent Dropout and Early School Leaving. OECD Education Working Papers, 53, pp. 1-64.
MacMillan, D., Ballo, I., Widman, K., Borthwick-Duffy, S., & Hendrick, I. (1990). Methodological problems in estimating dropout rates and the implications for studying dropouts from special education. Exceptionality, 1, 29-39.
McNeal, R. B. (1997). High school dropouts: A closer examination of school effects. Social Science Quarterly, 78 (1), 209-222.
Mincer, J. (1958). Investment in human capital and personal income distribution. Journal of the Political Economy, 4, 281-302.
MoEYS and ADB. (2012). Project Design Recommendations: School-based Enrichment Program. Phnom Penh: Aisan Development Bank.
MoEYS. (2005). Education Stategic Plan 2006-2010. Phnom Penh: Ministry of Education Youth and Sports.
MoEYS. (2013). Education Statistics and Indicators 2012/2013. Ministry of Education Youth and Sport.
MoEYS. (2010). Education Strategic Plan 2009-2013.
33
MoEYS. (2014). Education Strategic Plan 2014-2018.
Montmarquette, C., Viennot-Briot, N., & Dagenais, M. (2007). Dropout, school performance and working while in school. The Review of Economics and Statistics, 89 (4), 752-760.
NIS. (2006). CSES 2006 field manual. Cambodia Socio-Economic Survey .
Pan, J. (2010). The Effects of Child Labor on School Attendance in Cambodia. Graduate School of Arts & Sciences: Georgetown University.
Peter, D. (2008). Child labour, education and health: A review of the literature. International Labour Office .
Phoumin, H., & Fukui, S. (2006). Cambodian Child’s Wage Rate, Human Capital and Hours Worked Trade-Off: Simple Theoretical and Empirical Evidence for Policy Implications. Kobe University Graduate School of International Cooperation Studies Working Paper.
Rajani, S., Jennifer, S., & Karen, T. (2011). Dropout Trend Analysis for Cambodia – School Dropout Prevention. the United States Agency for International Development (USAID).
Ravallion, M., & Quentin, W. (1999). Does Child Labor Displace schooling Schooling? Evidence on Behavioral Responses to an Enrollment Subsidy. Washington DC: World Bank.
Ricardo, S., Kwame, A., Jo, W., & Frances, H. (2010). School Drop out: Patterns, Causes, Changes and Policies. School of Education and Social Work, University of Sussex, the Centre for International Education. UNESCO.
Roderick, M. (1993). The Path to Dropping Out: Evidence for Intervention. Westport, CT: Auburn House.
Rumberger, R. W. (2011). Dropping out: Why students drop out of high school and what can be done about it. Cambridge MA: Harvard University Press.
Rumberger, R. W. (2001). Why Students Drop Out of School and What Can Be Done. K-12 Dropouts and Graduation Rates, UCLA.
Suliman, E. D., & El-Kogali, S. E. (2002). WHY ARE THE CHILDREN OUT OF SCHOOL?:Factors Affecting Children’s Education in Egypt. 1-33.
Texas Associaiton of School Boards. (2007). The Dropout, Completion and Graduation Rates. Taxas.
34
Thurlow, M. L., Sinclair, M. F., & Johnson, D. R. (2002). Students with disabilities who drop out of school: Implications for policy and practice. Issue Brief, 1 (2).
UNESCO. (2009). Education Indicators Technical Guidelines.
UNICEF, ILO, & WB. (2006). Children’s Work in Cambodia: A Challenge for Growth and Poverty Reduction. Understanding Children's Work: An Inter-Agency Research Cooperation Project .
Appendix
Table 7: Average Distance to Nearest School in Kilometer
distance to nearest primary school
distance to nearest lower secondary school
non-dropout 1.12 2.83
dropout 1.20 3.08
Total 1.12 2.85
Source: Author’s calculation from CSES 2003-2012
Table 8: Overall pooled effect sizes by outcome and intervention type
35
Source: Shari et al (2013), page 46
Note: *p<0.1, **p<0.05 and ***p<0.001
36